Virtues in War: A Review of “A Cooperative Species” by Bowles and Gintis

Humans are not born selfish. But why? Bowles and Gintis suggest that Darwin was right: altruism is good the group. But this leads them to the troubling suggestion that virtue evolved in times of war.

A Cooperative Species by Samuel Bowles, Herbert Gintis | Paper Plus

After E. O. Wilson published his Sociobiology in 1975, the term became soon associated with explanations of violence, rape, and sex differences. Similarly, Darwinism was seen by many economists as a justification for the assumption that humans are self-interested. Human biology was no field for humanistic optimism.

The tide started to change around the turn of the century. More and more, scholars became interested in the higher realms of human nature. A central figure in this shift was Herbert Gintis, one of the few high profile academics who proudly identifies as a sociobiologist. In 2001, Gintis proclaim that “the central task of sociobiology is to explain the origins of prosocial emotions”. A decade later, the A Cooperative Species was published: a monumental book by Gintis and his long-time collaborator, economist Samuel Bowles.

A very ambitious oeuvre, A Cooperative Species seeks to explain the evolutionary origins of human cooperativeness in all its forms. For such a daunting topic, the book is very, very comprehensive. In this aspect, it has very few rivals. I can only think of Tomasello’s Natural History of Human Morality and Krebs’ Origins of Morality. Even in this company, A Cooperative Species is unique in its strong footing in game theory and economics. Indeed, I would recommend this book especially to readers trained in mathematical subjects – and specifically, to any economists who still feels an affinity with the myth of a purely self-interested Homo economicus.

Personally, I will return to the book many times to come. However, the heavily mathematical argumentation makes the prospects of returning to this beast somewhat agonizing. I think the authors should have published the formal modelling in appropriate journals, and then used to book to explain the conclusions to a wider readership. The decision to include all modelling in the book makes it very challenging for those, like me, who don’t have a background in economics or population genetics. For casual readers, this makes the book downright unreadable. At some points, I almost get the impression that the authors are more interested in showcasing their scholarly capabilities, rather than in making their work accessible to the wider intellectual culture. Every one to their taste, I guess, but the outcome is a tragically diminished readership. This is a shame, for their argument deserves much attention.

As for the argument: Bowles and Gintis explore three claims. One is general in its nature and somewhat uncontroversial. The second and third are more specific and much more controversial.

First claim: Humans are not only driven by self-interest

The general claim is that people are not only selfish. We have a genuine taste for the welfare of others. We feel a genuine aversion to those who hurt others. We like to cooperate and we dislike cheaters. In their words:

“Because we are convinced that most people enjoy cooperating at least in some situations and dislike people who do not, the task we will set for ourselves is not that typically addressed by biologists and economists, namely to explain why people cooperate despite being selfish. Rather, we seek to explain why we are not purely selfish—why the social preferences that sustain altruistic cooperation are so common.”

This broad thesis tries to purge the topic from the many hangovers of 20th Century biology and economics. For the authors claim, rightly I think, that these hangovers are pernicious – and most probably false. Contra Dawkins, selfish genes do not mean that humans are “born selfish”. Contra the tragedy of commons, humans do not always exploit common resources in the absence of sanctions. And contra naive interpretations of game theory, even the prisoner’s dilemma does not always ask for defection.

The authors amass a convincing array of data to support this thesis. They do so well. Using experimental evidence from various cultures around the world, they conclude that although “behaviours are highly variable across groups” it is nevertheless true that “not a single group approximated the behaviours implied by the self-interest axiom”. Cultures differ in their style of prosociality. But they all have a style.

Second claim: Group selection explains the evolution of human altruism

More controversial is their account of the evolution of our cooperative nature. Here, Bowles and Gintis argue that the traditional explanations of kin selection and reciprocal altruism fall short. Kindness for strangers is not just an accidental by-product of our capacity to love for our family members. Nor is it always a form of “enlightened self-interest”, where we secretly expect returns in the future.

This specific set of criticisms is comprehensive and well-grounded in empirical data. For example, they agree that kin selection would technically allow altruism to evolve in groups, but only in groups significantly more homogenous than a typical hunter-gatherer band. Similarly, reciprocal altruism requires groups way smaller than a hunter-gatherer band.

Instead, they favour a form of group selection, coupled with some gene-culture coevolution. The idea is that although sacrificing for others is costly to the individual, it helps the flourishing of the group. Therefore, groups with more altruists outcompete other groups.

This idea goes back to Darwin himself. In On the Origin of Species, he wrote:

“it hardly seems probable that the number of men gifted with such virtues [as bravery and sympathy] … could be increased through natural selection, that is, by the survival of the fittest.”

Instead, he saw sympathy and other virtues as a product of evolution for the good of the group.

“Those communities which included the greatest number of the most sympathetic members would flourish best and rear the greatest number of offspring.” 

Nevertheless, group selection -theories were largely discredited in the mid-20th Century. Most of this criticism was targeted at very naive interpretations of group selection. Bowles and Gintis are part of an attempt to bring a more nuanced picture back.

This is perhaps their greatest contribution. Admittedly, they do build on pre-existing work. For many decades now, David Sloan Wilson has lead the argument that group selection can be understood as part of the Dawkinsian gene-eye view of evolution. Groups are not alternative units of selection; they are vehicles for genes to propagate. But Wilson’s general argument had an achilles heel: it’s critics agreed that group selection can work in principle, but they maintained that, in reality, it is unlikely to have affected human ancestors in any significant way.

This is where Bowles and Gintis come in. They argue that the worry is misplaced. Instead, their models demonstrate that very plausible parameters from human prehistory would allow an “altruism allele” to spread in human populations. This, then, is their second claim: the most altruistic forms of human cooperativeness evolved by group selection in the Pleistocene.

But in the details of these models lurks the third, and most controversial claim: Bowles and Gintis claim that group selection for altruism was possible because human prehistory was rigged with warfare. It was killing, not kindness, which created the evolutionary corridor for altruism to evolve.

The authors are not celebrating war. Yet they write:

“We initially recoiled at this unpleasant and surprising conclusion. But the simulations and the data on prehistoric warfare tell a convincing story.”

This, then, is the third claim of their book.

Third claim: Group selection for altruism was made possible by high levels of warfare

Before jumping into the details, it is worth looking at the basic idea. Roughly, the idea is that during war, bravery and sacrifice become important for the survival of the group. Groups with a lot of altruists were more likely to survive than during other times. So, groups with high levels of egoists were annihilated, and so, absolute amount of egoists in the human gene pool diminished slowly.

Psychologically, the basic idea is captured in the sentiment of a “band of brothers”. Indeed, Bowles and Gintis quote a set of Israeli studies, which show that citizens behave more cooperatively in economic games during times of war. Unfortunately, the kind of virtue required in war is one of parochial altruism, instead of universal kindness. Whether this is a virtue of a vice of the model is difficult to say. For humans are certainly biased towards their ingroup, but like Bowles and Gintis admit, altruism does not actually “stop at the borders”.

But let us return to the evolutionary logic. As said, nothing in the gene-eyed perspective of modern evolutionary theory makes the group selection theory impossible in principle. The mathematics is clear. Group selection can work in principle. In many cases, it works better than traditional models of kin selection. So, the question is whether this could have happened in human prehistory.

Bowles and Gintis use parameters of group size and warfare levels, which they deem plausible in the light of archeological and ethnographic evidence. The guesses are questionable, but the approach is wonderful. Few if any scholars have gone to similar lengths in constraining formal game-theoretic models which such detailed empirical considerations. Of course, the amount of assumptions is gargantuan. But so will it be in any formal argument. For my money, they show that human group selection is a solid hypothesis. It might be false, but no obvious factor makes it impossible.

Does this mean that only war can make group selection work? Here, the authors misrepresent their models in a surprising way. For after arguing in (unconvincing) detail that the Pleistocene was extremely violent, they admit that for some of their models to work, only a 3,6% mortality from warfare is needed. This does require the existence warfare, which is itself questionable, but the figure is very low. It certainly does not mean that war was omnipresent. And so, their long background argument about an extremely militarized Pleistocene is unnecessary.

Furthermore, the authors admit that repeated climatic changes, something we know occurred in the Pleistocene, would probably do an equally good job at creating the required parameters for group selection to take place. Later, Bowles and Gintis even admit to using simulation settings that are “quite unfavourable to the evolution of altruism”. For example, they neglect the potential effects of selective emigration and mate-choice. A “realistic” model in such an unrealistic setting is difficult to interpret.

In other words, the authors present a very specific argument, but with repeated acknowledgements that, even if the details are incorrect, the basic argument would still hold. In some sense, this is good. It further supports the plausibility of their second claim: group selection could explain human altruism. But in this vein, their argument should not read: “a lot of warfare, therefore altruism”. Rather, it should read “even a bit of warfare or climatic change or whatever, therefore altruism”.

I think the latter formulation would have served the field better. This is evident in critical replies to the work, like Kim Sterelny’s Pleistocene Social Contract. Sterelny challenges Bowles and Gintis’ assumptions of Pleistocene warfare. Here, I am inclined to side with Sterelny. A telling example is rock art. Bowles and Gintis themselves use prehistoric rock art as an evidence for ancient warfare amongst the aborigines. But tellingly, there is not a single example of Pleistocene rock art depicting warfare. As Sterelny admits, the data we have is “patchy” and “might be misleading”. But he concludes, rightly I think, that “the evidence we have does not suggest a militarized Pleistocene.”

This is all well and good. But as said, the Bowles-Gintis model does not necessarily require a highly militarized Pleistocene. So although I remain sceptical of their third claim, I would urge as to take their central argument more seriously. The general argument is that, as Darwin suggested, group selection might well have played an important role in human prehistory.

Other issues: Chimpanzees, bonobos, and the “altruism allele”

Besides their treatment of warfare, I found some difficulty in accepting the authors’ overly adaptationist method. They tend to assumes that any quirk in human behaviour must be a direct response to a specific evolutionary pressure. This story is convenient for mathematically-minded thinkers like Bowles and Gintis, who feel a warm affection towards the formal elegance of the Price equation. But biologically, it neglects the holistic nature of many evolutionary pathways.

Take the chimpanzee-bonobo example. When compared to chimps, bonobos show much more altruistic behaviour, such as food sharing towards strangers. But this is probably not due to a direct evolutionary pressure to share food with strangers. Of course, it might be so. But this is not the most plausible explanation.

The most plausible explanation is the self-domestication –hypothesis, where the social selection in bonobo groups leads to an evolutionary premium on a friendly temperament. The process seems pretty straightforward. Bonobo foraging style allows females to band together, and therefore, for the females to gang up on aggressive alphas. Therefore, the bonobo gene pool has moved away from aggressive traits and towards more docile traits. (Indeed, a docile bonobo male has more offspring, on average, than an aggressive chimpanzee alpha.) Like dogs evolving from wolves, this causes against a holistic evolutionary pathway towards increased friendliness.

This is relevant for two reasons.

First, it shows that prosocial traits can evolve in a holistic fashion. This sheds doubt on the strictly adaptationist framework, where the quest is to find a mythical “altruism allele”.

Second, we should recall that Bowles and Gintis see warfare “as a powerful evolutionary force that paradoxically might account for social solidarity among humans”. But no such link exists in our closest relatives. Quite the opposite: chimpanzees have evolved under elevated risk of war, whereas bonobos live remarkably peacefully. Yet it is bonobos, not chimpanzees, who excel in social solidarity. Admittedly, the authors could argue that it is not warfare itself, but warfare coupled with some traits that humans have and chimps lack. Fair enough. But at first glance, the chimpanzee-bonobo comparison casts a decent shadow of scepticism on the war-altruism hypothesis.

(I am curious to see if this issue will come up in future communication between Bowles, Gintis and self-domestication scholars like Richard Wrangham or Brian Hare. I sure hope it does!)


Overall, A Cooperative Species is a rather technical and challenging work, which contributes to the field in two was: It collects heaps of interesting data into one neat package, and it presents strong modelling in favor of group selection -theories of human altruism. Therefore, it’s is an essential contribution to the science of human sociality. Its heavy betting on a wartorn prehistory remains controversial. But we should remember that this assumption is as critical as it appears.

Notes and links

The Dawkins quote is from the opening chapter of the 1976 edition of the Selfish Gene. He later apologized for this “rogue claim”.

The capacity of many societies to avert the “tragedy of the commons” without state or private ownerships has been demonstrated by Nobel laureate Elinor Ostrom.

For work on group selection by David Sloan Wilson, see his classic book Unto Others (co-authored with Elliot Sober) or his shorter and more recent book, Does Altruism Exist. For a shorter piece, see this article, co-authored with Edward O. Wilson (not a relative).

For D.S. Wilson’s model of group selection, which is based on assumptions of high inter-group emigration, see this paper (co-authored with Lee Dugatkin). This is an interesting model, for the assumption of high immigration is the diametrical opposite to the high group-fidelity in Bowles and Gintis’ models.

For a landmark paper on the mathematical convenience of group selection over kin-selection, see this paper by Martin Nowak, Corita Tarnita and E. O. Wilson.

For archeological and anthropological evidence suggesting that war did not exist in the Pleistocene, see work by Douglas Fry, e.g. his book The Human Potential for Peace, and this 2013 article in the magazine Science, co-authored by Patrik Söderberg (a fellow Finn!). These works have been popularized by science writer John Horgan amongst others.

For work on bonobo self domestication, see Wrangham and Hare’s co-authored paper. For the work on self-domestication in humans, see their respective books: Goodness Paradox, and Survival of the Friendliest. If you prefer an article-length summary, see Hare’s 2017 paper.

See also my abridged review in

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: